Saturday, February 4th 2017
First Intel Processor with AMD Radeon Graphics Within 2017
Back in December, it was reported that Intel could license AMD's GPU technology for integration into its future processors. The whispers are growing louder, with Hard|OCP editor Kyle Bennett (who broke the original December story), reporting that the first product of this collaboration could be out within 2017. According to Bennett, posting on the Hard|OCP Forums, the first Intel product with AMD Radeon GPU IP could be a special processor with an AMD Radeon GPU die, and a CPU die based on the "Kaby Lake" micro-architecture.
Bennett further adds that the Radeon-enhanced Intel processor could be a multi-chip module (MCM) with the Radeon GPU die being separate from the CPU die, it won't be an on-die component such as Intel's own HD Graphics solution. This could also mean that AMD will supply nearly-finished dies to Intel, likely manufactured at its own trusted fabs (Global Foundries or TSMC), and not hand over sensitive designs over to Intel's fabs. The product could be an entry-mid range product, which means Intel is trying to aim for the value consumer segment, and not necessarily the workstation crowd. Bennett concludes that one could expect more collaboration between Intel and AMD over graphics IP in the future.
Source:
HardOCP
Bennett further adds that the Radeon-enhanced Intel processor could be a multi-chip module (MCM) with the Radeon GPU die being separate from the CPU die, it won't be an on-die component such as Intel's own HD Graphics solution. This could also mean that AMD will supply nearly-finished dies to Intel, likely manufactured at its own trusted fabs (Global Foundries or TSMC), and not hand over sensitive designs over to Intel's fabs. The product could be an entry-mid range product, which means Intel is trying to aim for the value consumer segment, and not necessarily the workstation crowd. Bennett concludes that one could expect more collaboration between Intel and AMD over graphics IP in the future.
77 Comments on First Intel Processor with AMD Radeon Graphics Within 2017
AMD has 10% of the market. Even if Ryzen was much faster than Kaby Lake, and it isn't, Intel's influence and brand recognition would be enough to keep Ryzen at one quarter of the market, best case scenario. We saw something like that 12 years ago with Athlon64 vs Pentium 4.
So, why fight to go from 10% to 25%, when you can get payed for the other 75%-90% of the market? OK maybe you will not go to that 25% and stay at 15%-20%, but the income from selling GPUs to Intel will be significant, probably much higher than what AMD will make from an extra 5%-10% of the midrange market where APUs will be playing ball.
Then there are other areas where AMD will gain. We already have seen AMD's gains from the consoles. It wasn't money, but the changes in game development.
The consoles where enough to change the game market. Nvidia tried with PhysX and GameWorks, but it couldn't stop the console ports tsunami. So, while games in the past where written for Nvidia hardware, something that was catastrophic for AMD, in driver's stability and performance, now they are written and optimized for GCN. The results can be seen this last year, with AMD cards not only continuing in gaining against Nvidia's cards in benchmarks, but also Nvidia having problems with it's driver's stability. Not to mention DX12 and Vulkan.
With Intel going GCN, GCN becomes like the x86 for graphics cards. Nvidia's next architecture should be a marvel of engineering for developers to not ignore it. Not to mention that FreeSync and other AMD techs will become the de facto standards over night.
Nvidia is a threat to both AMD and Intel. But mostly Intel, because Intel is the huge company here, Intel is the company that will lose the most. An alliance between AMD and Intel to throw out Nvidia of the desktop market, will benefit both tremendously. In the future the war between ARM and x86 platforms will only intensify and Nvidia can become a huge asset for the ARM camp, both in the gaming market and the AI market. If AMD and Intel manage to limit Nvidia's expansion, or even more, manage to turn Nvidia's fast growth to a recession, they will have better chances to defend against the ARM platform.
But second, If Intel want to make it in their own they will have to redesign it actually as the manufacturing process is different between these fabs, so they will have to redesign it to fit their process and also they will need to optimise it further as it's a completely new design to them. in this case the cost it self will be very high as it will add more time to fab it, and a lot of engineering work hours not just from Intel, but also from AMD.. Intel might as well disclose sensitive manufacturing secrets to AMD's engineer for such collaboration.
Edit: Yup, NVIDIA/Intel licensing agreement ends March 31, 2017. Intel can keep producing GPUs perpetually, but they get nothing new from NVIDIA after that date. Intel GPUs are effectively no longer getting upgrades so in models where they want good graphics, they're just buying AMD chips.
That article was written in 2015 and at the time, 30% of NVIDIA's profit came from this Intel licensing deal. I think Intel sees what NVIDIA is doing and don't like funding them anymore. Buy chips in bulk from AMD as demand requires ends up being a lot cheaper anyway.
Additionally, it provides a buffer from an anti-trust lawsuit if Intel resumes shady tactics to keep Zen at bay. They can point to their balance sheet and say "we paid AMD $x million dollars for AMD GPUs!" A judge would have to take that into consideration when fining Intel.
TL;DR: This hurts NVIDIA hugely and helps AMD a little directly (big purchaser of chips) but a lot indirectly (this is going to put Radeon Settings on a lot of prebuilts which is going to give Radeon mindshare it desperately needs).
I just meant I don't know but that could work too
dont believe me
google it or ask politely for a link :)
Additionally, NVIDIA's revenue is going to fall from $5 billion to $3.5 billion. They have to tighten their belt so starting a new, costly R&D project in a highly competitive market isn't exactly going to excite investors.